The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes a skill-report.json file that contains a pre-computed security verdict claiming the skill is safe and that any findings are false positives. This is a self-referential technique (Category 8e) used to influence security analyzers.
Ingestion points: skill-report.json file content and metadata fields.
Boundary markers: Absent for the metadata processing phase.
Capability inventory: File system writes for test files and configuration; command execution via npm install and test runners.
Sanitization: None provided for the metadata fields or user-supplied testing requirements.
[COMMAND_EXECUTION]: The instructions in SKILL.md prompt the agent to execute shell commands for installing dependencies (npm install --save-dev jest ...) and running tests (npm run test). While standard for the stated purpose, these are active command execution patterns.

backend-testing